Forecasting global temperatures: Missing the point? The consequences of the hiatus

Abstract

We use singular spectrum analysis techniques to discriminate the underlying signals within the HadCRUT4 global surface temperature record. Our analysis identifies a multidecadal oscillation (related to natural oscillations) and a secular trend (assumed to be representative of anthropogenic-induced warming) as the two main signals within the temperature record. Most current generation global circulation models (CMIP5) do not reproduce the multidecadal oscillation and fail to capture the present observed temperature hiatus in their simulations. Therefore, it is unlikely that these models can correctly forecast the temperature evolution during the coming decades. Statistical forecasts based on the analyzed secular trend and the multidecadal oscillations are indeed capable of reproducing the observed hiatus and generally result, in comparison to CMIP5 forecasts, in much lower temperature increases for 2100 of only about +0.39°C [-0.47-2.46] assuming a “business as usual” scenario. Either the global energy budget uncertainty is still too large or the increased radiative forcing does rather lead to accelerated warming of other parts of the climate system as the ocean or the cryosphere.